WIDAR3.0: WiFi-based Activity Recognition Dataset

To stimulate the development of wireless sensing, we produce this wifi-based activity recognition dataset to the community. This dataset includes the Channel State Information (CSI) collected from commodity Wi-Fi devices for gestures and Body-coordinate Velocity Profile (BVP) calculated by the algorithms descried in the Widar3.0 paper. The hand gesture dataset consists of 258K instances of data samples with a duration of 8,620 minutes and from 75 domains. Also included in this repository is the gait recognition dataset related to the GaitID paper. The gait recognition dataset consists of 22K instances of data samples from 11 participates. Please stay tuned for further updates.References:Yue Zheng, Yi Zhang, Kun Qian, Guidong Zhang, Yunhao Liu, Chenshu Wu, Zheng Yang, "Widar3.0: Zero-Effort Cross-Domain Gesture Recognition With Wi-Fi", ACM MobiSys, 2019. DOI:https://doi.org/10.1145/3307334.3326081Yi Zhang, Yue Zheng, Guidong Zhang, Kun Qian, Chen Qian, Zheng Yang, "GaitID: Robust Wi-Fi Based Gait Recognition", Springer WASA, 2020. DOI:https://doi.org/10.1007/978-3-030-59016-1_60

搜集汇总

背景与挑战

背景概述

WIDAR3.0是一个基于商用WiFi网卡收集的大规模手势识别数据集，包含258K个手势实例和75个领域的数据，提供RSSI、CSI原始信号以及DFS、BVP两种高级特征。该数据集总时长8620分钟，未来还将扩展步态识别等更多活动识别数据。